Padraig Gleeson (NPP) awarded EPSRC Open Fellowship

The fields of neuroscience and AI have traditionally taken different approaches to developing, studying and utilising “brain-inspired” network models. In his EPSRC fellowship, Dr. Gleeson will develop a framework that will enable researchers from neuroscience, AI and cognitive science to create and exchange computational models across all of these fields.

The EPSRC Open Fellowship will allow Dr. Padraig Gleeson of the Department of Neuroscience, Physiology and Pharmacology to significantly expand his work on the development of advanced computational models of neurons and neuronal circuits. Until now, his work has focussed primarily on models that incorporate the physiological and anatomical detail present in their biological equivalents. He has spearheaded the development and uptake of the NeuroML model description language, a community standard for specifying models in computational neuroscience, as well as the Open Source Brain platform, which provides the infrastructure to enable open source, collaborative development of biophysically detailed models, and the open sharing and analysis of experimental data. 

There is an increasing need in computational neuroscience, however, to incorporate recent developments being made in the field of Artificial Intelligence (AI), and the neural network architectures being investigated there. Huge advances have been made in software frameworks for creating and applying machine learning models, and many of these carry out tasks (visual processing, navigation, decision making) which are actively studied by neuroscientists in biological networks. However, there are many practical impediments to the exchange of research outputs between these domains, related to different formalisms, methods and software infrastructure used, as well as the varied backgrounds of researchers involved.

A solution to this was outlined in a recent NeuroView article in Neuron: Integrating model development across computational neuroscience, cognitive science, and machine learning, Gleeson et al. 2023. To facilitate this convergence, Dr. Gleeson’s 5 year fellowship will involve the development of a specification language, MDF (Model Description Format), which will allow models across all these fields to be expressed and exchanged, facilitating the sharing of insights from multiple perspectives on intelligent information processing in networks. The language will integrate with neuronal models in NeuroML, machine learning models in PyTorch and TensorFlow and many others, including models of cognitive processes, in collaboration with Prof. Jonathan D. Cohen at Princeton Neuroscience Institute. 

In addition, the fellowship will enable Dr. Gleeson’s research group to contribute to the OpenWorm project, which aims to create a cell by cell computational model of the nematode C. elegans to investigate how a complete nervous system can control the behaviour of a simple animal in a simulated 3D environment. His fellowship includes a unique “Plus Component”, allowing Dr. Gleeson to dedicate a portion of his time to outreach, community engagement and promotion of open science practices. Connecting students and researchers interested in these different fields will also continue through his involvement with the UCL NeuroAI initiative.